Predicting Therapy Outcomes in Patients With Stress-Related Disorders: Protocol for a Predictive Modeling Study

Area of Science:

• Mental Health
• Psychiatry
• Computational Medicine

Background:

• Cognitive behavioral therapy (CBT) is effective for adjustment and exhaustion disorders, but predictors of treatment response remain unclear.
• Identifying these predictors can refine assessment and treatment strategies for stress-related disorders.
• Combining traditional prediction methods with machine learning can enhance understanding of treatment response.

Purpose of the Study:

• To evaluate predictors of treatment response in stress-related disorders using traditional methods.
• To model treatment outcomes via machine learning, integrating interpretability with pattern recognition.
• To compare traditional and machine learning approaches for predicting treatment success.

Main Methods:

• Analysis of data from a randomized controlled trial of internet-delivered CBT versus an active control for adjustment/exhaustion disorders (N=300).
• Pooled data analysis incorporating sociodemographic, clinical, self-rated, and cognitive variables.
• Application of univariate logistic regressions, ablation studies, and machine learning classifiers (elastic net logistic regression, random forest, SVM, AdaBoost) with 70/30 train/test split and 5-fold cross-validation.

Main Results:

• Hypothesized predictors include younger age, education, baseline symptom severity, treatment credibility, and prior sickness absence.
• Machine learning models are expected to outperform a majority class prediction baseline.
• Anticipated balanced accuracy of ≥67% for machine learning models, indicating clinical utility.

Conclusions:

• This study addresses the limited research on treatment outcome predictors for stress-related disorders.
• Findings may facilitate personalized treatments for adjustment and exhaustion disorders, improving clinical practice.
• The dual approach could promote larger studies and clinical implementation of machine learning for precision mental health.